[OpenVPN home] [Date Prev] [Date Index] [Date Next]
[OpenVPN mailing lists] [Thread Prev] [Thread Index] [Thread Next]
Web openvpn.net

[Openvpn-users] performance numbers part 2

  • Subject: [Openvpn-users] performance numbers part 2
  • From: kt o <krt2941@xxxxxxxxxxx>
  • Date: Thu, 13 Dec 2007 20:14:19 -0800
  • Importance: Normal

Given the feedback, I went back and ran the network performance tests again.
I borrowed another physical machine to use as the client (instead of a virtual machine).  This setup won't be used in a real deployment - I just needed to find some boxes to test with.  But the eventual openVPN server will be windows OS based (don't have any spare boxes to run linux/bsd/etc)

OpenVN Server - AMD Athlon64 3800+ (2 Ghz), windows XP
OpenVPN Client - AMD opteron 2210 (1.8 ghz), windows 2003
The openVPN configs are from the sample / HOWTO.  Encryption is the default (BF-CBC).  And TLS is used (shouldn't be a factor either way right?)

Again, using netperf 2.4.4.  Values are throughput in 10^6 bits/sec.

Client -> Server: (i.e. client runs netperf and server runs netserver)
No VPN: 90.27, 88.63, 89.25, 88.46
*Hamach VPN: 66.70, 66.79, 66.29, 66.66
Dev Tun, plaintext, no-LZO: 48.82, 49.62, 48.30, 49.70
Dev Tun, plaintext, LZO: 53.98, 54.95, 53.94, 54.84
Dev Tun, encrypted, no-LZO: 37.79, 39.71, 38.07, 37.36
*Dev tun, encrypted, LZO: 50.69, 50.68, 51.38, 50.62
Dev Tap, encrypted, LZO: 44.66, 44.29, 44.65, 44.00

Server->Client: (i.e. client runs netserver and server run netperf)
No VPN: 89.40, 90.70, 89.80, 91.00
*Hamachi: 59.12, 62.84, 61.56, 63.50
Dev Tun, plaintext, no-LZO: 63.70, 64.03, 61.06, 63.73
Dev Tun, plaintext, LZO: 63.11, 63.47, 64.14, 63.48
Dev Tun, encrypted, no-LZO: 50.51, 49.43, 50.41, 50.44
*Dev Tun, encrypted, LZO: 59.40, 59.51, 59.22, 59.49
Dev Tap, encrypted, LZO: 51.59, 51.33, 51.46, 51.62

Final thoughts:
For the netperf generated traffic (it says TCP STREAM TEST), Dev Tap is slower across the board in all permutations of encrypted X LZO.  So is proto TCP.

I don't know why Hamachi VPN fares better in the Client -> Server direction, whereas OpenVPN fares better in the Server->Client direction.  I would have thought it was CPU related, but then both VPNs should either perform better or worse for the same direction.

The result is that looking solely at Server->Client numbers, OpenVPN and Hamachi are quite even.  But looking at the Client->Server numbers shows OpenVPN roughtly 22% slower than Hamachi.  This seems too strange and I feel that my test setup is screwed up somehow - any thoughts?

The best games are on Xbox 360. Click here for a special offer on an Xbox 360 Console. Get it now!